• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Hardware Raid vs. Onboard Raid Controller

boomshakalaka

boomshakalaka

Weaksauce
Joined
Dec 18, 2007
Messages
77
Hi everyone,

I recently purchased a supermicro X8SI6-F with an on-board LSI SAS 2008 8-port SAS2 controller. However, reading some articles online some people consider "On-board raid" as software controllers? and that they do not perform as well as an add on third party card because they are only software based and thus use the CPU to controll the RAID. I always thought they were hardware controlled because there is a chip for them on the motherboard..:confused:

Supposedly the supermicro onboard LSI SAS 2008 8-port SAS2 controller is essentially a LSI 9211-8i RAID controller shipping with IR firmware.. so is this controller considered a hardware, hybrid, or software raid?

Any input is appreciated, Thank you.
 
Easiest way to tell if something is considered hardware RAID is whether or not it includes memory. With hardware RAID, the controller will handle parity calculations and take a load off the CPU. If you are doing something that doesn't use parity however (RAID 0/1 and the nested levels) then it won't really make a difference. What you have on the board is somewhere in the middle since it isn't hardware RAID and it isn't software based and will still perform better than software based RAID.
 
The SAS2008 chip is far superior to the typical peashooter fakeraid chips bundled on a lot of motherboards - for one its got 8 lanes x 6Gbps of bandwidth. That said, its not a XOR chip, that is its not capable of parity calculation meaning no raid4/5/6 and nested raid types.

the real question is what you want to do with it, not what everyone else's definition of software or hardware raid is. its good as an HBA driving an expander, or to do simple RAID0/RAID1/RAID10/JBOD.
 
odditory said:
The SAS2008 chip is far superior to the typical peashooter fakeraid chips bundled on a lot of motherboards - for one its got 8 lanes x 6Gbps of bandwidth. That said, its not a XOR chip, that is its not capable of parity calculation meaning no raid4/5/6 and nested raid types.

the real question is what you want to do with it, not what everyone else's definition of software or hardware raid is. its good as an HBA driving an expander, or to do simple RAID0/RAID1/RAID10/JBOD.
Click to expand...

Thinking of using this board with ESXi as follows:

2x74GB SATA (Raid1) for OS
3x1TB SATA (Raid 10) for VM storage
 
Should be more than sufficient for that, though you'll need a 4th drive for RAID 10.
 
What OS will you be running? On Windows the software RAID options are limited and you have to use the proprietary drivers; so this is driverRAID under Windows. But other OS use it as normal HBA like Linux/Solaris/BSD.
 
I have been playing with both the X8SI6-F and the LSI 9211-8i... really decent performance and it is identical between the two solutions (I've been getting +/- 3% results which is a fair margin of error).

If you aren't running RAID 6, this is going to be fine.
 
ESXi loves that SAS2008 chip so there won't be any issue there - my ESXi server is running on a X8SI6 board.
 
hw raid is just a dedicated cpu that runs some software. There is not really a difference between hardware raid and software raid.

The trend today is that the computer's cpu runs all software. The hardware card is ditched, and all software runs on the cpu. Just like ZFS.

The cpus get more and more powerful, soon you dont need dedicated cpus any longer. The cpu can run all software - just like ZFS. ZFS is open source, and more tested, and safe than hw raid - which might corrupt your data. ZFS is easy to patch. Just get rid of all raid solutions and let ZFS handle all discs. That is the easiest, safest and cheapest solution.
 
brutalizer said:
hw raid is just a dedicated cpu that runs some software. There is not really a difference between hardware raid and software raid.

The trend today is that the computer's cpu runs all software. The hardware card is ditched, and all software runs on the cpu. Just like ZFS.

The cpus get more and more powerful, soon you dont need dedicated cpus any longer. The cpu can run all software - just like ZFS. ZFS is open source, and more tested, and safe than hw raid - which might corrupt your data. ZFS is easy to patch. Just get rid of all raid solutions and let ZFS handle all discs. That is the easiest, safest and cheapest solution.
Click to expand...
You have no idea what you are talking about once again. We don't need you preaching ZFS. Did you even read what the original poster wrote? They couldn't even use ZFS if they wanted to. You just seem to be looking for any thread with "hardware RAID" in it as an opportunity to preach ZFS. Do us all a favour and either go away or make some quality posts. :rolleyes:
 
Blue Fox said:
They already listed the OS.
Click to expand...
Well i can't find it. And who's 'they'? I thought the OP was likuid.silence and he began his thread with "i" not "we".

I only can read that ESXi is gonig to be the Hypervisor; but i can't find what OS the OP wants to run underneath the hypervisor. I don't know ESXi well, but it does mean the target OS must support the controller as well right; since it gets physical access to the hardware?
 
sub.mesa said:
Well i can't find it. And who's 'they'? I thought the OP was likuid.silence and he began his thread with "i" not "we".

I only can read that ESXi is gonig to be the Hypervisor; but i can't find what OS the OP wants to run underneath the hypervisor. I don't know ESXi well, but it does mean the target OS must support the controller as well right; since it gets physical access to the hardware?
Click to expand...
"They" can be used in a singular gender neutral form when referring to someone in third person, in this case, the OP. As for ESXi, if it supports a device, the OS doesn't need to unless you want to pass a device through with VT-d last time I checked.
 
Blue Fox said:
Easiest way to tell if something is considered hardware RAID is whether or not it includes memory. With hardware RAID, the controller will handle parity calculations and take a load off the CPU. If you are doing something that doesn't use parity however (RAID 0/1 and the nested levels) then it won't really make a difference. What you have on the board is somewhere in the middle since it isn't hardware RAID and it isn't software based and will still perform better than software based RAID.
Click to expand...

To add to this, a full blown hardware RAID card will have local ECC memory, and a battery backup unit to guard against data corruption in the event of a power failure. A system such as the one you describe will use ECC memory (I'm guessing the CPU is a Xeon) and you can achieve redundant power with an UPS and/or dual power supplies.

As to the ESXi question, I'd assume that since ESXi is essentially a super-thin OS itself, and has its own Hardware Compatibility List, the guest OSs would not need support for the physical hardware, but would be presented with virtualised devices by the hypervisor, in the same way that VMware Player & Workstation and Oracle VirtualBox do.

EDIT:

BlueFox, you beat me while I was typing the post. :p
 
But what about storage, isn't there a huge overhead if you're building a fileserver and 'emulating' all storage devices? I thought modern hypervisors (like Xen?) would share REAL hardware, requiring a compatible kernel on all guest OS. This way, you would have direct access to the hardware and can use all normal functions and guest OS drivers, without having a huge penalty on virtual device emulation.

But i'm very green with virtualization, and haven't played with Xen for real yet. Not sure how it compares to ESXi either. So feel free to update my understanding of this. :D
 
Not really. It wouldn't be very wise for the guest OSs themselves to see the hardware directly unless you pass it through with VT-d/IOMMU. How else would you move a VM from one system to another? Windows for example isn't very fond of booting off different HBAs. Things such as high-availability wouldn't really be possible either and that's critical in many applications.
 
I find that the single hard disk overhead (when the storage is in the host) is large with vmware-server, vmware-player, esxi, xen and virtualbox. However kvm is not so bad if you use vritio disk. To me most of these basically turn a WDC black into performing like a 5400 RPM laptop hard drive from 4 years ago. And I am talking about higher end host systems. > 3.0 GHz i7s with 6+ GB of ram and 3GB of ram dedicated to the guest. I have not tried raid arrays or SSDs yet.

BTW, this opinion/observation was made by months of testing and frustration trying to get 64 bit windows guests to run under 64 bit linux and vise-versa with the usage case being developing medical imaging applications that have 2 to 3 million lines of c++ source code.
 
Last edited:
You must log in or register to reply here.
Back
Top